If you want to use your model to answer questions concerning the structure, function and stability of the mature protein, you should base your model on the mature sequence of the protein. Be aware of the fact that in nature, the signal sequence is removed before translation has completed and the protein can fully fold. Whether the presence of the signal peptide interferes with the structure prediction depends on the method used and, for homology modelling, on the degree of homology to the best available template. The programs might simply not model the signal peptide as it is not present in high homology templates representing experimental structures, direct the template selection towards a template less suitable for the prediction of the mature protein, or produce a prediction for the signal peptide moiety that has little to do with reality, as the signal peptide would probably be highly aggregation-prone in solution, as its natural environment is the translocation pore and the lipid membrane.

Annemarie Honegger Thank you very much for the detailed explanation!